1. Field of the Invention
This invention generally relates to thin film transistor (TFT) fabrication and, more particularly, to a method for extending the laser pulse length during the annealing and crystallization of silicon films.
2. Description of the Related Art
Polycrystalline silicon (poly-Si) material is typically used as the active layer of poly-Si TFTs in the fabrication of active-matrix (AM) backplanes. Such backplanes can be used to fabricate AM liquid-crystal displays (LCDs) and can be also combined with other display technologies, such as organic light emitting diodes (OLEDs).
Poly-Si material is typically formed by the crystallization of initially deposited amorphous Si (a-Si) films. This process can be accomplished via solid-phase-crystallization (SPC), by annealing the a-Si films in a furnace at appropriate temperature and for sufficiently long time. Alternatively, laser annealing can also be used to achieve the phase transformation.
These conventional crystallization techniques are applied to a substrate in a manner that yields uniform poly-Si film quality throughout the substrate area. With the SPC method, long annealing times are required, such as 30 hours at 600 C. On the other hand, the laser annealing method has an advantage of greater throughput, as it requires less time for crystallization than the SPC method.
The poly-Si material crystallized using the conventional laser annealing method consists of many grains, with each grain being surrounded by grain boundary. The size of these grains is typically one micron. However, the typical channel length of TFT is approximately 2 to 30 microns. Therefore, it is inevitable that the channel regions of TFT contain several grain boundaries. These grain boundaries act as an electron (hole) trap, degrading TFT performance characteristics and reliability.
Recently, a new laser annealing technique has been developed that allows for significant flexibility in the process techniques, permitting controlled variation in resulting film microstructure. This technique relies on lateral growth of Si grains using very narrow laser beams, that are generated by passing a laser beam through a beam-shaping mask, and projecting the image of the mask to the film that is being annealed. The method is called Laser-Induced Lateral Growth (LILaC).
FIG. 1 is a diagram illustrating the LILaC process (prior art). The initially amorphous silicon film is irradiated by a very narrow laser beamlet, with typical widths of a few microns (i.e. 3-5 xcexcm). Such small beamlets are formed by passing the original laser beam through a mask that has open spaces or apertures (see FIG. 2), and projecting the beamlets onto the surface of the annealed Si-film.
FIG. 2 is a conventional laser annealing mask (prior art). Returning to FIG. 1, the sequence of images 1 through 4 illustrates the growth of long silicon grains. A step-and-repeat approach is used. The shaped laser xe2x80x9cbeamletxe2x80x9d (indicated by the 2 parallel, heavy black lines) irradiates the film and then steps by a distance smaller than half of the width of the slit. As a result of this deliberate advancement of each beamlet, grains are allowed to grow laterally from the crystal seeds of the poly-Si material formed in the previous step. This is equivalent to laterally xe2x80x9cpullingxe2x80x9d the crystals, as in zone-melting-crystallization (ZMR) method or other similar processes. As a result, the crystal tends to attain very high quality along the xe2x80x9cpullingxe2x80x9d direction, in the direction of the advancing beamlets. This process occurs simultaneously at each slit on the mask, allowing for rapid crystallization of the area covered by the projection of the mask on the substrate. Once this area is crystallized, the substrate moves to a new (unannealed) location and the process is repeated.
FIG. 3 is a pictorial representation of a system to accomplish the optical projection and the step-and repeat process (prior art). In order to improve the throughput, longer lateral growth length and larger laser beam dimensions (laser irradiation area per pulse) are required.
Some poly-Si materials formed through the LILaC process have a highly periodical microstructure, where crystal bands of specific width are separated by high-angle grain boundaries. Within the crystal bands, low-angle boundaries are observed with a frequency of occurrence dependent upon certain specifics of the crystallization process, such as film thickness, laser fluence (energy density), pulse duration, and the like. TFTs fabricated on such poly-Si films demonstrate very good characteristics, as long as the direction of conduction is parallel to the direction of the in-crystal low-angle boundaries.
TFTs with greater electron mobility can be fabricated if the substrate crystallization characteristics can be made more isotropic. In other words, the TFT performance depends upon the angle between the main crystalline growth direction, the direction parallel to the laser scanning axis, and the TFT channel. This is due to the formation of sub-boundaries within the crystal domains. The surface roughness at the xe2x80x9chardxe2x80x9d grain boundaries, at the edges of the crystal bands/domains, can be significant. This surface roughness prohibits the reduction of the gate insulator thickness, which is one critical step for scaling down the device geometry for future applications. Further, not all these processes can be location controlled. Therefore, by chance only, depending upon the relative size of the crystal domain and the TFT channel length, certain TFTs will not include grain-boundaries in their active area (channel), whereas other TFTs will include one or more boundaries in their active areas. This kind of non-uniformity is highly detrimental for critical-application TFTs where uniformity of characteristics is more essential than absolute performance.
It would be advantageous if the throughput of TFTs fabricated using the LILaC process could be improved.
It would be advantageous if the lateral growth length of crystallized areas could be lengthened using the LILaC process.
It would be advantageous if the laser energy could be used more efficiently in the LILaC process.
The present invention is based upon on relationship explored between laser pulse length and the maximum lateral growth rates. In order to achieve higher throughput in the LILaC process, longer lateral growth length and/or larger laser beam dimensions are needed. As a general trend, the lateral growth length (LGL) becomes longer as the laser energy density (fluence) increases. The maximum fluence level is limited by agglomeration and maximum laser energy of equipment.
An extension of the pulse duration can bring an associated increase in the LGL, but only if the laser fluence is appropriately increased. For the same laser fluence, a longer pulse duration will not necessarily result in longer LGL. This effect stems from the balance between additional energy input in the film (as a result of longer heating) and additional output from the system as a result of increased heat losses. The balance between these two phenomena ultimately determines whether the film will remain molten for a longer time, thus permitting longer lateral growth to occur.
At high levels of fluence, some damage to the substrate may be caused. Hence, the advantage of achieving longer LGL must be balanced against the possible substrate damage and/or chance of contamination. At the same laser fluence, thinner films demonstrate longer LGL than thicker films, based on the thermal balance between input energy and heat losses through the substrate. With thicker Si-films, a longer pulse duration helps achieve longer LGL, only if the energy density is also significantly increased. However, as stated above, such high fluence may have detrimental effects on the integrity of the substrate. Therefore, the benefit of the longer pulse duration cannot be viewed independently from the fluence requirements. The higher fluence requirement affects two areas: (a) throughput, as a higher fluence demands a smaller beam area (beam dimensions) for a given laser power output, and (b) substrate integrity, as a higher fluence may affect the substrate and, as a minimum, requires a thicker base coat thickness be used. Longer pulse durations are used more advantageously in the crystallization of thin films. However, if the a-Si thickness is too thin, agglomeration very can easily and the poly-Si quality deteriorates. One optimal a-Si thickness is in the range of approximately 20 to 60 nanometers (nm). The xe2x80x9cextentxe2x80x9d of pulse extension is a matter of balancing the benefit of heat input against inadvertent heat losses. Data indicates that the best results are achieved using a pulse width in the range between 70 and 120 nanoseconds (ns), preferably around 100 ns.
Accordingly, an efficient extended pulse laser annealing method is provided. The method comprises: supplying a substrate with a thickness; selecting an energy density; selecting an extended pulse duration; laser annealing a substrate region; in response to cooling the substrate region, crystallizing the substrate region; and, efficiently extending the lateral growth of crystals in the substrate region.
When the substrate has a thickness of approximately 300 xc3x85, the energy density is selected to be in the range of 400 to 500 millijoules pre square centimeter (mJ/cm2). The pulse duration is selected to be in the range between 70 and 120 nanoseconds (ns). More preferably, the pulse duration is selected to be in the range between 90 and 120 ns. Most preferable, the pulse duration is approximately 100 ns. Then, efficiently extending the lateral growth of crystals in the substrate region includes laterally growing crystals at a rate of approximately 0.029 microns per nanosecond.
Additional details of the above-described method, and details of the method as applied to thicker substrates are provided below.